• 0
• 0
• 0

Summary

This podcast explores the use of deep learning to predict Bitcoin price direction. Several deep learning models, including CNN-LSTM, LSTNet, and TCN, are compared to an ARIMA model as a benchmark. Feature selection methods are employed to optimize model performance. The results show that the combination of Boruta feature selection with the CNN-LSTM model outperforms others, achieving an accuracy of 82.44% and generating extraordinary annual returns in simulated investment strategies. The study highlights the potential of predictive models in Bitcoin trading, while acknowledging limitations and suggesting areas for future research.

Duration: 16 minutes (00:15:50)
Publishing date: 2024-12-11; Adapted; Copyright Year: 2024. Copyright Statment: —

• Business & Economics
• Science & Technology